Beruflich Dokumente
Kultur Dokumente
Experimental Gerontology
journal homepage: www.elsevier.com/locate/expgero
A R T I C LE I N FO A B S T R A C T
Section Editor: Emanuele Marzetti Background: Although, previous studies have reported a positive association between hemoglobin levels and
Keywords: bone mineral density (BMD), the majority of the studies were limited in patients with chronic hypoxemic
Hemoglobin conditions and findings concerning the association among non-anemic populations are inconclusive. We aimed
Bone mineral density to examine the association between hemoglobin levels and BMD in non-anemic healthy adults.
Osteopenia Methods: This cross-sectional study included 3626 non-anemic men and women aged ≥ 60 years who partici-
Osteoporosis pated in the Korean National Health and Nutrition Examination Survey (KNHANES). The BMD of the lumbar
spine and both femurs was measured by dual-energy X-ray absorptiometry (DXA). Participants with T-score for
BMD < −1.0 SD were defined as having low BMD. The odds ratios (ORs) and 95% confidence intervals (CIs) for
low BMD were calculated using multiple logistic regression analyses across sex-specific hemoglobin quartiles.
Results: The prevalence of low BMD gradually decreased in accordance with hemoglobin quartiles in both sexes.
Compared with the group in the lowest quartile, the OR (95% CI) for low BMD in the lumbar spine was 0.78
(0.54–0.93) for men and 0.67 (0.50–0.93) for women after adjusting for age, BMI, tobacco smoking, alcohol
intake, physical activity, walking difficulty, household income, total calorie intake, calcium intake, iron intake,
25(OH)D, alkaline phosphatase, and parathyroid hormone levels. However, these positive associations were not
found in femur after adjusting for the same co-variables.
Conclusions: Hemoglobin levels were inversely associated with low BMD in lumbar spine among non-anemic
adults.
☆
Sources of support for research: None.
⁎
Corresponding author at: Department of Family Medicine, Yonsei University College of Medicine Gangnam Severance Hospital, 211 Eonju-ro, Gangnam-gu, Seoul,
Republic of Korea.
E-mail address: ukyjhome@yuhs.ac (Y.-J. Lee).
https://doi.org/10.1016/j.exger.2019.110706
Received 17 June 2019; Received in revised form 1 August 2019; Accepted 19 August 2019
Available online 20 August 2019
0531-5565/ © 2019 Elsevier Inc. All rights reserved.
H.-S. Kim, et al. Experimental Gerontology 126 (2019) 110706
BMD have been based on non-axial peripheral BMD measurement ratio of weight (kg) to height2 (m2). BMI was categorized into three
methods, whereas the International Society for Clinical Densitometry groups, namely, underweight (BMI < 18.5 kg/m2), normal (BMI
(ISCD) and the WHO recommend central DXA in axial bone such as 18.5–24.9 kg/m2), and obese (BMI ≥ 25.0 kg/m2) according to the
lumbar spine, total hip, or femoral neck (Organization, 1968). Likewise, Korean Society for the Study of Obesity. Participants were asked about
the study populations used in most of the previous studies on this area lifestyle behaviors, including cigarette smoking and alcohol consump-
of research have been limited to Western populations with chronic tion. Smoking status was categorized as a current smoker (a person who
hypoxemic medical conditions such as chronic obstructive pulmonary smokes cigarettes daily) or not-current smoker (a person who has never
disease (Rutten et al., 2013), renal failure (Taal et al., 1999), and an- smoked or who smoked in the past but who does not currently smoke
emia (Sarrai et al., 2007). cigarettes). Alcohol consumption was assessed into two categories
In the present study, we investigated the association between he- based on how often participants consumed any type of alcohol, namely,
moglobin levels within the normal range and central bone mineral current drinker (a person who drinks alcohol, even once a month) or
density as measured by DXA in apparently healthy Korean men and not-current drinker (a person who drinks less than once a month).
women based on the 2009–2010 Korean National Health and Nutrition Participants completed the International Physical Activity
Examination Survey (KNHANES). Questionnaire (IPAQ) to determine the frequency of physical activity.
Household income was classified into four quartiles from the lowest to
2. Materials and methods the highest. Dietary intake was assessed by administrating a Food
Frequency Questionnaire (FFQ) using a single 24-h recall method.
2.1. Study population Before testing, all participants were instructed to maintain their usual
dietary habit. Total calorie intake, calcium intake, and iron intake were
This study was based on data obtained from the 2009 and 2010 calculated by Can-Pro 2.0, computerized nutrient intake assessment
KNHANES, a nationally representative survey conducted by the Korean software developed by the Korean Nutrition Society. If participants
Ministry of Health and Welfare. The target population of the survey were being treated for any disease, they were asked for data regarding
included non-institutionalized civilians in Korea who were over one their diagnosis as well as a list of medications currently being taken.
year of age. Sampling units consisted of households selected through a Completed questionnaires were reviewed by trained staff and entered
stratified, multistage, probability-sampling design based on geographic into a database. Participants were also divided into two groups ac-
area, sex, and age group using household registries. Sampling weights, cording to the degree of physical activity, either difficulty in walking or
which were representative of the probability of being sampled, were no difficulty in walking.
assigned to each participant to ensure that the results represented the Serum hemoglobin levels were measured using an SLS hemoglobin
overall Korean population. Participants completed a four-part survey (no cyanide) method with an XE-2100D instrument (Sysmex, Tokyo,
that consisted of a health interview, health behavior questionnaire, a Japan). Serum 25(OH)D concentrations were measured with a radio-
health examination, and a nutrition questionnaire. immunoassay (Diasorin Inc., Stillwater, MN, USA) using a γ-counter
In this study, 5608 participants aged ≥ 60 years old were included (1470 Wizard; PerkinElmer, Turku, Finland). The coefficient of varia-
in the 2009 and 2010 KNHANES. We excluded participants who met at tion of control material was ≤7.6% for low level and ≤5.8% for high
least one of the following criteria: those with a history of anemia, is- level, respectively. The central testing institute has participated in the
chemic heart disease, stroke, thyroid, hepatobiliary, chronic renal dis- international Vitamin D External Quality Assessment Scheme (DEQAS).
ease, or cancers; those with hemoglobin level < 13 g/dL in men, < Results of the DEQAS proficiency testing showed less than ± 2.0 stan-
12 g/dL in women; those with missing data for BMD (T-score), he- dard deviation index (SDI), except the SDI range from −2.40 to +2.87
moglobin, alkaline phosphatase, parathyroid hormone, and 25(OH)D due to random error in 2010, without negative or positive trend. Serum
levels; those receiving hormone replacement therapy; those taking vi- parathyroid hormone (PTH) levels were measured using the N-tact PTH
tamin and mineral supplements. Following these exclusions, a total of assay with LIAISON (Diasorin Inc., Stillwater, MN, USA), which is a
3626 Koreans (1673 men, 1953 women) were in included in the final chemiluminescence immunoassay method.
analysis after exclusions. The KNHANES was approved by the The BMD (g/cm2) measurements of central skeletal sites (femoral
Institutional Review Board of the Korea Centers for Disease Control and neck, and lumbar spine) were obtained using DXA (QDR 4500A;
Prevention (Approval numbers: 2009-01CON-03-2C and 2010-02CON- Hologic Inc., Waltham, MA, USA). BMD measurements were performed
21-C). In addition, this study complied with the ethical principles of the according to the 2007 ISCD recommendations. The DXA instruments
Declaration of Helsinki. were calibrated as previously described (Schoeller et al., 2005), and
reference values were obtained using this method (Kelly et al., 2009).
2.2. Data collection DXA calibration was performed by measuring phantoms, which are
reference standards, on a daily basis. The results were analyzed using
For the 2009 and 2010 KNHANES, citizens were informed that they standard techniques and Korean Society of Osteoporosis and Hologic
had been randomly selected as a household to voluntarily participate in Discovery software (version 13.1). L1–4 values were chosen for the
a nationally representative survey conducted by the Korean Ministry of determination of lumbar spine BMD. When these specific vertebrae
Health and Welfare, and that they had the right to refuse participation were not suitable for analysis due to compression fracture, degenerative
in accordance with the National Health Enhancement Act supported by changes or any other reason, the calculation of BMD excluded the af-
the National Statistics Law of Korea. All participants gave written in- fected vertebrae. Participants with a BMD T-score of < −1.0 standard
formed consent to participate in the study. The Korea Centers for deviation (SD) were placed in the low BMD group.
Disease Control and Prevention also obtained written informed consent
to use blood samples from participants for further analysis. The health 2.3. Statistical analysis
examinations, which were performed in 2009 and 2010, included a
medical history, physical examination, a questionnaire about health- Because hemoglobin levels differ significantly by sex, hemoglobin
related behaviors, anthropometric measurements, and biochemical in- quartiles were categorized separately as follows: Q1, ≤14.2; Q2,
dices. 14.4–14.8; Q3, 15.0–15.6; and Q4, ≥15.7 g/dL for men and Q1, ≤12.7;
Physical examinations were performed by trained medical staff Q2, 12.8–13.2; Q3, 13.2–13.8; and Q4, ≥13.9 g/dL for women. In this
following standardized procedures. Body mass and height were mea- survey, the sample weights were constructed for sample participants to
sured to the nearest 0.1 kg and 0.1 cm, respectively, in light indoor represent the Korean population by accounting for the complex survey
clothing, without shoes. Body mass index (BMI) was calculated as the design, survey non-response and post-stratification. Therefore, in
2
H.-S. Kim, et al.
Table 1
The demographic and biochemical characteristics of the subjects according to hemoglobin quartiles.
Q1 ≤ 14.2 Q2 14.3–14.8 Q3 14.9–15.6 Q4 ≥ 15.7 P-valuea Post-hocb Q1 ≤ 12.7 Q2 12.8–13.2 Q3 13.3–13.8 Q4 ≥ 13.9 P-valuea Post-hocb
3
2nd quartile 27.1 ± 2.6 24.1 ± 2.8 27.1 ± 2.5 23.1 ± 2.5 23.7 ± 2.5 25.2 ± 2.3 24.2 ± 2.2 26.0 ± 2.4
3rd quartile 17.7 ± 2.1 24.5 ± 3.0 24.3 ± 2.3 24.6 ± 3.2 16.1 ± 2.3 20.9 ± 2.3 16.5 ± 1.9 19.7 ± 2.5
4th quartile 14.8 ± 1.9 16.4 ± 2.6 26.5 ± 2.7 23.2 ± 3.5 13.2 ± 1.9 17.1 ± 2.4 15.6 ± 2.2 13.1 ± 2.6
Total calorie intake (g/day) 2015 ± 48 2077 ± 55 2162 ± 45 2077 ± 45 0.124 – 1550 ± 28 1594 ± 49 1503 ± 30 1454 ± 32 0.081 –
Daily calcium intake (g/day) 497.9 ± 19.1 577.6 ± 24.8 564.7 ± 19.3 547.2 ± 19.7 0.036 a,b 390.2 ± 13.1 421.3 ± 20.5 401.2 ± 16.3 379.1 ± 13.9 0.351 –
Daily iron intake (g/day) 15.8 ± 0.7 17.1 ± 0.9 17.3 ± 0.7 16.6 ± 0.7 0.368 – 12.5 ± 0.4 14.0 ± 0.7 12.5 ± 0.5 13.8 ± 1.0 0.159 –
25(OH)D (ng/mL) 20.5 ± 0.5 20.7 ± 0.5 21.4 ± 0.5 20.7 ± 0.5 0.405 – 18.1 ± 0.4 18.5 ± 0.6 18.0 ± 0.5 18.6 ± 0.4 0.610 –
Alkaline phosphatase (IU/L) 251.8 ± 4.1 241.7 ± 5.4 236.6 ± 3.4 249.4 ± 4.7 0.020 b 256.2 ± 4.7 254.5 ± 4.2 263.4 ± 4.2 273.2 ± 4.5 0.017 c, e
PTH (pg/mL) 68.3 ± 2.0 66.2 ± 1.8 66.3 ± 1.5 65.8 ± 1.5 0.758 – 70.3 ± 1.7 70.6 ± 1.9 71.0 ± 1.8 71.9 ± 1.7 0.893 –
T-score
Femur neck −0.991 ± 0.053 −0.887 ± 0.066 −0.703 ± 0.055 −0.648 ± 0.062 < 0.001 b,c −1.706 ± 0.054 −1.634 ± 0.065 −1.533 ± 0.055 −1.461 ± 0.047 0.008 b,c,e
Spine −0.727 ± 0.079 −0.689 ± 0.087 −0.586 ± 0.063 −0.493 ± 0.0.075 0.023 c,e −1.877 ± 0.058 −1.868 ± 0.071 −1.688 ± 0.065 −1.539 ± 0.064 0.010 c,e
Fig. 1. Prevalence of low BMD according to hemoglobin quartiles in men and women. Post hoc analysis of weighted chi-squared test for mean between-group
difference (a. Q1 versus Q2; b. Q1 versus Q3; c. Q1 versus Q4; d. Q2 versus Q3; e. Q2 versus Q4; f. Q3 versus Q4).
4
H.-S. Kim, et al. Experimental Gerontology 126 (2019) 110706
Fig. 2. Odds ratios 95% confidence intervals for low BMD according to hemoglobin quartiles in men (A) and women (B). Multiple logistic regression analyses were
conducted after adjusting for age, body mass index, tobacco smoking, alcohol intake, physical activity, walking difficulty, household income, total calorie intake,
calcium intake, and iron intake, serum levels of 25-hydroxyvitamin D, alkaline phosphatase, and parathyroid hormone.
composed of fat cells. Red blood cells, platelets, and most subtypes of stimulation of fibroblast proliferation, increased osteogenic potential,
leukocytes are produced in the red marrow (Gurevitch et al., 2007). In and stimulation of adipogenesis (Malladi et al., 2007). A recent rodent
this way, the age-related conversion of red to yellow bone marrow in study showed that hypoxia blocks the growth and differentiation of
the axial skeleton may explain how hemoglobin levels are inversely osteoblasts while strongly stimulating osteoclast formation (Arnett,
associated with low BMD (Schellinger et al., 2004; Schwartz et al., 2010). Because the majority of the body's oxygen supply is transported
2013). Moreover, spine bone is composed of a 3:1 ratio of trabecular by hemoglobin, the inverse association between hemoglobin levels and
and cortical bone, whereas this ratio is 1:1 in the femoral bone. The low BMD in healthy elderly individuals may be explained by the hy-
trabecular bone turnover rate is known to be about eight- to ten-fold poxia mechanism.
higher than that of cortical bone, because trabecular bone contains There are some limitations of the present that should be considered
most of the body's red bone marrow, i.e. hematopoietic stem cells when interpreting its findings. First, this was a cross-sectional study by
(Clarke, 2008). In this regard, lumbar spine was thought to more sen- design and thus it was not possible to establish a causal relationship
sitively reflect the loss of red marrow than femoral neck in the present between hemoglobin levels and BMD. Further prospective studies with
study. a larger number of participants over a longer period of time are war-
Secondly, relative hypoxia accompanied by decreased hemoglobin ranted to confirm the cause and effect relationship of hemoglobin levels
levels may have an adverse effect on bone metabolism via the direct and BMD. Second, hemoglobin levels in this study were only measured
action of reduced pO2 and an indirect effect on bone formation and once, and thus it was not possible to determine whether acute or brief
resorption cells (Bozec et al., 2008; Knowles and Athanasou, 2009; episodes of low hemoglobin levels were responsible for the observed
Utting et al., 2010). Oxygen tension has long been considered to be correlations. Despite these potential limitations, the inverse association
important in skeleton homeostasis. Published data regarding the effects between hemoglobin levels and BMD reported in this paper are based
of hypoxia on bone marrow mesenchymal stem cells include on nationally representative data and may be useful for the clinical
5
H.-S. Kim, et al. Experimental Gerontology 126 (2019) 110706
setting. Moreover, we believe that the present study is the first to study. Osteoporosis international: a journal established as result of cooperation be-
specifically examine the relationship between hemoglobin levels and tween the European Foundation for Osteoporosis and the National Osteoporosis
Foundation of the USA 21, 1761–1768. https://doi.org/10.1007/s00198-009-
BMD in a non-anemic healthy elderly population. 1131-7.
Kelly, T.L., Wilson, K.E., Heymsfield, S.B., 2009. Dual energy X-ray absorptiometry body
5. Conclusions composition reference values from NHANES. PLoS One 4, e7038. https://doi.org/10.
1371/journal.pone.0007038.
Knowles, H.J., Athanasou, N.A., 2009. Acute hypoxia and osteoclast activity: a balance
In conclusion, hemoglobin levels were inversely associated with low between enhanced resorption and increased apoptosis. J. Pathol. 218, 256–264.
BMD the lumbar spine among non-anemic elderly individuals. https://doi.org/10.1002/path.2534.
Korkmaz, U., Korkmaz, N., Yazici, S., Erkan, M., Baki, A.E., Yazici, M., Ozhan, H.,
Ataoglu, S., 2012. Anemia as a risk factor for low bone mineral density in post-
Author contributions menopausal Turkish women. European journal of internal medicine 23, 154–158.
https://doi.org/10.1016/j.ejim.2011.11.009.
Laudisio, A., Marzetti, E., Pagano, F., Bernabei, R., Zuccala, G., 2009. Haemoglobin levels
H-S Kim designed the study, wrote and edited the manuscript. Y-J
are associated with bone mineral density in the elderly: a population-based study.
Lee and H-S Lee did statistical analyses and interpreted the data. H-M Clin. Rheumatol. 28, 145–151. https://doi.org/10.1007/s10067-008-0998-6.
Park, and Y-J Lee contributed to the discussion and revised the manu- Leng, S., Chaves, P., Koenig, K., Walston, J., 2002. Serum interleukin-6 and hemoglobin
script. Y-J Lee, as the corresponding author, coordinated the study, as physiological correlates in the geriatric syndrome of frailty: a pilot study. J. Am.
Geriatr. Soc. 50, 1268–1271.
interpreted the data, contributed to the discussion, and wrote the Malladi, P., Xu, Y., Chiou, M., Giaccia, A.J., Longaker, M.T., 2007. Hypoxia inducible
manuscript. All authors read and approved the final version. factor-1alpha deficiency affects chondrogenesis of adipose-derived adult stromal
cells. Tissue Eng. 13, 1159–1171. https://doi.org/10.1089/ten.2006.0265.
Onal, E.D., Usluogullari, A., 2012. Anemia and osteoporosis: causal association or epi-
Declaration of competing interest phenomenon? European journal of internal medicine 23, e117. https://doi.org/10.
1016/j.ejim.2012.03.010.
No potential conflict of interest was reported by the authors. Organization, W.H., 1968. Nutritional Anaemias: Report of a WHO Scientific Group
[Meeting Held in Geneva From 13 to 17 March 1967].
Papadimitriou, N., Tsilidis, K.K., Orfanos, P., Benetou, V., Ntzani, E.E., Soerjomataram, I.,
Acknowledgements Kunn-Nelen, A., Pettersson-Kymmer, U., Eriksson, S., Brenner, H., Schottker, B.,
Saum, K.U., Holleczek, B., Grodstein, F.D., Feskanich, D., Orsini, N., Wolk, A.,
Bellavia, A., Wilsgaard, T., Jorgensen, L., Boffetta, P., Trichopoulos, D.,
We thank all those who conducted the KNHANES as well as the Trichopoulou, A., 2017. Burden of hip fracture using disability-adjusted life-years: a
civilians who participated in this survey. pooled analysis of prospective cohorts in the CHANCES consortium. Lancet Public
Health 2, e239–e246. https://doi.org/10.1016/s2468-2667(17)30046-4.
Reginster, J.Y., Burlet, N., 2006. Osteoporosis: a still increasing prevalence. Bone 38,
Funding information
S4–S9. https://doi.org/10.1016/j.bone.2005.11.024.
Rutten, E.P., Franssen, F.M., Spruit, M.A., Wouters, E.F., 2013. Anemia is associated with
This research did not receive any specific grant from funding bone mineral density in chronic obstructive pulmonary disease. Copd 10, 286–292.
agencies in the public, commercial, or not-for-profit sectors. https://doi.org/10.3109/15412555.2012.744390.
Salive, M.E., Cornoni-Huntley, J., Guralnik, J.M., Phillips, C.L., Wallace, R.B., Ostfeld,
A.M., Cohen, H.J., 1992. Anemia and hemoglobin levels in older persons: relationship
References with age, gender, and health status. J. Am. Geriatr. Soc. 40, 489–496.
Sarrai, M., Duroseau, H., D’Augustine, J., Moktan, S., Bellevue, R., 2007. Bone mass
density in adults with sickle cell disease. Br. J. Haematol. 136, 666–672. https://doi.
Arnett, T.R., 2010. Acidosis, hypoxia and bone. Arch. Biochem. Biophys. 503, 103–109. org/10.1111/j.1365-2141.2006.06487.x.
https://doi.org/10.1016/j.abb.2010.07.021. Schellinger, D., Lin, C.S., Lim, J., Hatipoglu, H.G., Pezzullo, J.C., Singer, A.J., 2004. Bone
Bozec, A., Bakiri, L., Hoebertz, A., Eferl, R., Schilling, A.F., Komnenovic, V., Scheuch, H., marrow fat and bone mineral density on proton MR spectroscopy and dual-energy X-
Priemel, M., Stewart, C.L., Amling, M., Wagner, E.F., 2008. Osteoclast size is con- ray absorptiometry: their ratio as a new indicator of bone weakening. AJR Am. J.
trolled by Fra-2 through LIF/LIF-receptor signalling and hypoxia. Nature 454, Roentgenol. 183, 1761–1765. https://doi.org/10.2214/ajr.183.6.01831761.
221–225. https://doi.org/10.1038/nature07019. Schoeller, D.A., Tylavsky, F.A., Baer, D.J., Chumlea, W.C., Earthman, C.P., Fuerst, T.,
Cesari, M., Pahor, M., Lauretani, F., Penninx, B.W., Bartali, B., Russo, R., Cherubini, A., Harris, T.B., Heymsfield, S.B., Horlick, M., Lohman, T.G., Lukaski, H.C., Shepherd, J.,
Woodman, R., Bandinelli, S., Guralnik, J.M., Ferrucci, L., 2005. Bone density and Siervogel, R.M., Borrud, L.G., 2005. QDR 4500A dual-energy X-ray absorptiometer
hemoglobin levels in older persons: results from the InCHIANTI study. Osteoporosis underestimates fat mass in comparison with criterion methods in adults. Am. J. Clin.
international: a journal established as result of cooperation between the European Nutr. 81, 1018–1025. https://doi.org/10.1093/ajcn/81.5.1018.
Foundation for Osteoporosis and the National Osteoporosis Foundation of the USA Schwartz, A.V., Sigurdsson, S., Hue, T.F., Lang, T.F., Harris, T.B., Rosen, C.J., Vittinghoff,
16, 691–699. https://doi.org/10.1007/s00198-004-1739-6. E., Siggeirsdottir, K., Sigurdsson, G., Oskarsdottir, D., Shet, K., Palermo, L.,
Clarke, B., 2008. Normal bone anatomy and physiology. Clinical journal of the American Gudnason, V., Li, X., 2013. Vertebral bone marrow fat associated with lower trabe-
Society of Nephrology: CJASN 3 (Suppl. 3), S131–S139. https://doi.org/10.2215/cjn. cular BMD and prevalent vertebral fracture in older adults. J. Clin. Endocrinol.
04151206. Metab. 98, 2294–2300. https://doi.org/10.1210/jc.2012-3949.
Duh, M.S., Mody, S.H., Lefebvre, P., Woodman, R.C., Buteau, S., Piech, C.T., 2008. Taal, M.W., Masud, T., Green, D., Cassidy, M.J., 1999. Risk factors for reduced bone
Anaemia and the risk of injurious falls in a community-dwelling elderly population. density in haemodialysis patients. Nephrology, dialysis, transplantation: official
Drugs Aging 25, 325–334. https://doi.org/10.2165/00002512-200825040-00005. publication of the European Dialysis and Transplant Association - European Renal
Gurevitch, O., Slavin, S., Feldman, A.G., 2007. Conversion of red bone marrow into Association 14, 1922–1928. https://doi.org/10.1093/ndt/14.8.1922.
yellow - cause and mechanisms. Med. Hypotheses 69, 531–536. https://doi.org/10. Utting, J.C., Flanagan, A.M., Brandao-Burch, A., Orriss, I.R., Arnett, T.R., 2010. Hypoxia
1016/j.mehy.2007.01.052. stimulates osteoclast formation from human peripheral blood. Cell Biochem. Funct.
Ha, Y.-C., 2016. Epidemiology of osteoporosis in Korea. J Korean Med Assoc 59, 836–841. 28, 374–380. https://doi.org/10.1002/cbf.1660.
Jorgensen, L., Skjelbakken, T., Lochen, M.L., Ahmed, L., Bjornerem, A., Joakimsen, R.,
Jacobsen, B.K., 2010. Anemia and the risk of non-vertebral fractures: the Tromso